Inversion Of Iron Oxide Abundance From Hyperspectral Data Based On Wavelet Packet Analysis

Reflectance spectroscopy is usually associated with mineral types and abundances and its subtle changes with the chemical composition of the mineral,which in turn can interpret spectral data.Iron oxide is widely distributed on the earth's surface,and its mineral forms such as hematite,gothite and jarosite.Field studies show that their enrichment zones usually distributed around ore deposits,such as porphyry copper deposits.Therefore,in geological environment exploration iron oxide is often used as an indicator for mineral exploration.In this study,35 rock samples with known iron oxide content were selected from the ASTER spectral library,and the original reflectivity spectra were analyzed by wavelet packet.The approximate signals and detailed signals obtained were used to fitting iron oxide content by Partial least squares regression(PLSR).The optimal fitting model was applied to estimate the abundance of iron oxide from the satellite hyperspectral of GF-5 AHSI and EO-1Hyperion remote sensing images of the East Tianshan Mountains of Xinjiang.The main achievements of this study are as follows:(1)The detailed signal obtained by wavelet packet analysis can well explain the absorption characteristics of iron oxide.Therefore,fitting the abundance of iron oxide by the detailed signals,which can improve the fitting accuracy to a great extent.(2)Detail signals decomposed in the laboratory spectrum amplify the absorption characteristics of iron oxide.However,detail signal corresponding to the image is noise,while the signal related to iron oxide is retained in the approximate signal of the image.(3)The fitting equation of iron oxide abundance based on the detail signals was applied to the hyperspectral image to inversion the iron oxide abundance in the study area,and the results showed that the iron oxide abundance was consistent with the spatial distribution law of lithology.(4)According to the information of hyperspectral dimension of hyperspectral remote sensing data,the iron oxide abundance was estimated.Which can be used to accurately distinguish and identify the geological bodies of different lithologies,small changes of lithology,dikes and small structures,showing obvious technical advantages.